AGGREGATING RESULTS IN SOCIAL LIFE CYCLE IMPACT
ASSESSMENT
4.1 Présentation de l’article
Le manuscrit qui suit présente le développement méthodologique proposé pour quantifier et agréger les performances de l’analyse sociale du cycle de vie. Il a été rédigé sous la supervision des professeurs Pierre Baptiste et Manuele Margni.
Il a été soumis le 25 août 2016 à l’International Journal of Life Cycle Assessment et approuvé pour publication le 6 février 2017 (doi : 10.1007/s11367-017-1280-4).
4.2 Manuscrit
4.2.1 Introduction
In recent years, social life cycle assessment (SLCA) has emerged as an approach to assess the positive and negative social aspects in the life cycle of a product, from raw materials extraction to final disposal (Benoit et al., 2010). Zamagni et al. (2011) and Vinyes et al. (2013) observed that there were few case studies in SLCA and as such and were claiming that others are required to evaluate its application. According to Benoit et al. (2010), SLCA should be used to increase knowledge, clarify choices and promote the improvement of social conditions in the life cycle of products.
The approach may also be used to inform the decision-making process (e.g. to choose between two product systems, support responsible sourcing and supplier selection, improve management systems, foster responsible investment decisions, improve product design and optimize existing processes). However, there is no evidence of the use of SLCA for decision support (Jørgensen, 2013).
SLCA evaluates how life cycle activities affect people (Dreyer et al.2006). These activities are assessed based on the behaviours of the companies that are involved in the product system. Thus, it measures the social and socio-economic impacts in a supply chain (Benoît and Mazijn, 2009). As in environmental LCA (ISO 14044, 2006), SLCA is applied in four phases: definition of the goal and scope of study, life cycle inventory analysis, life cycle impact assessment and results interpretation. The main difference between the environmental and social approaches is in the way that the potential impacts are assessed.
Goal and scope: The same system boundaries as the environmental dimension are traditionally applied in SLCA, as per Vinyes et al. (2013). We observe the same for the functional unit. To measure social performance, subcategory indicators related to each stakeholder dimension are considered, such as workers, local community, society, consumers and participants in the value chain (Macombe et al., 2013). They may be defined in this phase and then assigned to the relevant stakeholder group (Reitinger et al., 2011). UNEP defines 31 types of impact subcategory indicators that may be considered in SLCA studies (UNEP, 2013).
Life cycle inventory analysis: Two approaches are currently used to assess social impacts (Ugaya, 2015): performance reference points (PRP) and social pathways. The first approach (type I) provides a qualitative assessment of the social performance of a product system’s activity. Scores attributed to the classification levels of the subcategory indicators generally express the qualitative assessment. The second approach (type II) requires substantial improvement in the development of cause and effect chains (Ekener-Peterson and Finnveden, 2013) and was not further considered in this study.
Considering type I approach, for each subcategory indicator, the classification levels are defined considering a performance reference point and data are collected. This process is guided by the selection of the impact subcategories identified in the goal and scope of the study and the activity variable is used to define the contribution of each phase of the life cycle for the entire product system (Benoît and Mazijn, 2009).
The PRP approach is widely used in scientific literature and indicated in UNEP guidelines to establish the classification levels of the indicators. It provides a basic requirement situation for the indicator and is the basis of the definition of risk, consistent, proactive and engaged
behaviours (Benoît and Mazijn, 2009). However, there is no standard process to establish these classification levels, considering type I approach.
Russo Garrido et al. (2016) identified different type I approaches to assess the social performance of a product system: (i) assessment based on standards and best practices; (ii) assessment based on standards and best practices and the socioeconomic context of unit process; (iii) assessment based on expert judgment on a business’ compliance with standards; (iv) assessment based on researcher expert judgment on business’ activities; (v) assessment based on a business, sector or country’s position with regards to an average and (vi) assessment based on how the data associated with the social performance of a business or a sector compare to other alternative businesses/sectors.
Based on the assessment based on standards and best practices approach, Sanchez-Ramirez et al. (2014) defined a framework to establish the classification levels of their indicators. According to the authors, the first step is to define a basic requirement (compliance threshold), creating two classes: compliant and non-compliant.
The compliant class may be divided into two subclasses: consistent and proactive. A company is proactive if it promotes good practices beyond compliance. The authors divide non-compliance into two subclasses by adding a new analysis based on the context in which the company operates. This investigation provides four classification levels, A (proactive), B (consistent), C (non-compliant, negative operation context) and D (non-compliant, positive operation context). Thus, companies are evaluated based on these classification levels for each indicator.
Life cycle impact assessment : The quantitative performance is calculated based on companies’ behaviour analyses (Benoît and Mazijn, 2009). The impact score is measured by aggregating the indicators in logical groups related to the social issues of interests to stakeholders’ dimensions and decision makers (Parent et al., 2010). A linear scale score is traditionally associated with each classification level of the subcategory indicators. The use of the linear function when translating qualitative classification levels into cardinal scores is arbitrary and was never justified despite the fact that the approach is proposed in several papers, as Sanchez-Ramirez et al. (2014), Dreyer et al. (2010) and Vinyes et al. (2013). In addition, they propose the same score sets for all subcategory indicators. We also remark these two limitations in the scoring methods identified in the critical review about SLCA development realized by Chhipi-Shrestha et al. (2015).
Different methods exist to perform the aggregation. Dong and Thomas Ng (2015) propose a survey of decision-makers and society in general to obtain weighting factors to establish the importance of the subcategory indicators.
However, they argue that it is impossible to guarantee that an expanded group of stakeholders has sufficient knowledge to assess the importance of each indicator. Integrating the value judgment of SLCA experts should be considered. Hosseinijou et al. (2014) used the Analytic Hierarchy Process (AHP) to generate the weights between the subcategory indicators. This method, however, is hard to be understood and can lead to weighting factors not representing the value judgment of the SLCA group of experts.
Interpretation: The interpretation phase provides information to support the decision-making process (Jolliet et al., 2010; Vinyes et al. 2013). Considering SLCA’s current level of development, practitioners cannot measure social impacts but only describe social performance (Macombe et al. 2013).
Finally, we identified two weakness of SLCA that may diminish the accuracy of the results: • The method used to translate the qualitative classification levels into scores relies on an
arbitrary linearity assumption for all subcategory indicators. This implies equal distance between the classification levels when translated into cardinal values.
• Applying sophisticated MCDA methodologies to define the weighting factors is heavy and – if not well understood – has the risk to generate sets of weighting factors that don’t represent the value judgement of the decision makers and sLCA experts.
Jørgensen and Jørgensen (2010) understood that to improve the social conditions in a product’s life cycle, it is necessary to provide a more robust assessment. Addressing the scoring and weighting systems is an important gap in SLCA (Benoit et al. 2010). There is no method or guidelines for the scoring and aggregation procedures (Benoit-Norris et al. 2011)—the main objective of this research work.
We therefore propose : 1) a scoring approach to translate the qualitative classification levels of subcategory indicators into quantitative performance scores that goes beyond the implicit linearity assumption and accounts for a customized value function based on SLCA expert
judgment; 2) a simplified process to define the weighting factors based on SLCA expert judgment used in aggregation. We then tested our approach into a illustrative case study.
4.2.2 Methods
Figure 4.1 describes the approach for the quantitative assessment of the social performance of product systems. It is structured in four phases: i) qualitative assessment; ii) definition of value function; iii) aggregation and iv) results’ interpretation.
The contribution of this research is the definition of customized value functions (phase ii) and the development of a weighting scheme to aggregate the subcategory indicators into stakeholder dimension indicators (phase iii). Both contributions on phases (ii and iii) can be used to conduct SLCA studies based on type I approach when translating qualitative performances into social scores.
4.2.2.1 Qualitative assessment
To assess social performance, we selected the sub-category indicators based on the 31 indicators set out by UNEP (UNEP, 2013). This choice was made based on our best knowledge of a relevant set of indicators for this case study. We considered the Performance Reference Point technique (PRP) to establish four classification levels (A, B, C and D) for each subcategory indicator (Figure 4.2) based on the framework proposed by Sanchez Ramirez et al. (2014). This framework was chosen because it considers not only the reference point to establish the classification levels, but also the social context where companies are operating. However, for some subcategory indicators (forced and child labour) it is impossible to measure the performance of the subcategory using a PRP approach because it is difficult to obtain specific data. In these cases, we consider the Social Hotspot Data Base (SHDB) indicators. As such, each phase of the life cycle was evaluated based on these indicators, considering the respective classification levels of each one.
Figure 94.2 – Performance Reference Point (PRP) technique to establish the classification levels of each subcategory indicator (adapted from Sanchez-Ramirez et al., 2014)
4.2.2.2 Definition of value function
To go beyond the simplified and unjustified assumption of linearity when translating the A to D classification levels into a numerical performance score, we propose an approach to develop a customized value function to account for the complexity and diversity of the subcategory indicators assessment. It involves a three-step procedure based on the judgment of SLCA experts: i) define the SLCA expert group; ii) define the customized value function; and iii) quantify the social performance.
For each subcategory indicator, SLCA experts are asked to place the ordinal classification levels (A, B, C and D) in a cardinal 0-10 scale, considering the description of the indicators provided (Figure 4.3). We defined the value function based on the value judgment of a SLCA group of experts, because we assumed they have the appropriate knowledge to understand and evaluate the distances among the classification levels of each subcategory indicator. It is possible adopting
other scales of this exercise (for example, [-10; 10]). As such, the use of the principle of negative/positive effects can be implemented easily, adapting the scale of Figure 3 for this case. For each subcategory indicator, numerical scores are normalized and then an average value representative of the SLCA expert group is obtained for each A-D classification levels.
Figure 104.3 – Tool to transfer the A-D ordinal classification levels of each subcategory indicator into a cardinal 0-10 scale scoring system
Interpolating the average values, three types of value function shapes to score the classification levels are possible: linear, concave and convex. The first is the shape currently considered in SLCA studies. For this shape, the distances among the classification levels are assumed to be equal, meaning that the scores of the classification levels increase linearly. The second is concave : the compliance threshold provides a higher score as compared to the linear shape. Finally, in the convex shape, the compliance threshold provides a lower score as compared to the linear shape. These results normalized into 0-1 are illustrated in Figure 4.4.
The proactive (classification level = A), consistent (classification level = B), non-compliant with negative operation context (classification level = C) and non-compliant with positive operation context (classification level = D) behaviors may also be scored differently on a cardinal scale depending on the subcategory indicator (Figure 4).
Figure 114.4 – Value function translating A-D ordinal classification levels into cardinal performance scores based on SLCA expert judgment specific to each subcategory indicator
The customized value functions as per Figure 4 are used by the SLCA practitioner to translate the qualitative assessment from PRP into a quantitative performance for each subcategory indicator. For each case study, different subcategory indicators can be adopted. The SLCA experts link ordinal scoring to the classification level of each subcategory indicator. This procedure is case- specific.
4.2.2.3 Aggregation procedure
We propose a two-level aggregation procedure. The first level aggregates the performances of all the life cycle stages for each criterion and considers the system reference flows and activity variable. For our illustrative case study, we considered the number of workers as activity variable
for the three stakeholder dimensions. We considered the same procedure as Sanchez-Ramirez et al. (2016) to calculate the activity variable.
The reference flows (RF) are defined based on the functional unit of the system, which represents the flow of material needed to provide the functional unit. In our simplified case study, we did not consider the impact of byproducts for the social dimension, but it can be considered in real- case application.
The amount of workers and the production capacity of the company are considered to establish the activity variable (eq. 1). A part-time worker was considered as half (0.5) working hours of a full-time worker. AC! = Wft! + (Wpt!∗ 0.5) PC! ∀i ∈ I eq. 1 Where,
AC!: Activity variable of the life cycle stage i (number of equivalent workers to produce an
output unit (liter) of the life cycle stage i) Wft!: Number of full − time workers
Wpt!Number of part − time workers
PC!Production capacity (output unit liter of the life cycle stage i)
I: set of life cycle stages
The amount of workers for the functional unit of each unit process given by (eq.2):
AW! = AC!∗ RF! eq.2
Where,
the functional unit − 1 liter of biodiesel)
RF!: Reference flow of the life cycle stage i (quantity of material necessary to produce the functional unit − 1 liter of biodiesel)
The aggregation factor (amount of workers) makes it possible to generate a profile of the product system in each subcategory indicator that aggregates all life cycle stages and is required to aggregate the performance scores at a higher level (stakeholder level, for example) to reduce the complexity for the decision-maker.
In order to go beyond the simplified and implicit assumption of equal weighting factors between the subcategory indicators, we propose the 2nd level aggregation, based on the importance granted to the indicators. The weights are obtained based on SLCA expert judgment. A hierarchical structure to establish the weights of the subcategory indicators is proposed.
The SLCA experts are asked to attribute points to each subcategory indicator within a given stakeholder dimension. The sum of the points equals 100 and is attributed based on the importance of the SLCA experts grant to each criterion in each stakeholder dimension. A set of weighting factors is obtained and aggregates the answers of the different SLCA experts.
We defined the value function based on the value judgment of the SLCA group of experts because we assumed that they are more familiar with social issues than decision-makers. As such, they are more able to evaluate the importance of each subcategory indicator to the stakeholder dimensions. However, decision-makers can be considered in this case if they have enough knowledge in SLCA to participate of this process.
For each stakeholder dimension, the weight of each subcategory indicator is defined as per eq. 3.
W!" = S!" S!" ! ∀ x ∈ X eq. 3 Where,
W!": Weight of subcategory indicator y of stakeholder x
X: Set of stakeholders
Y : Set of subcategory indicators
4.2.2.4 Results interpretation
In the last phase, the SLCA results of the different product systems obtained using customized and linear value functions are compared. We analyzed the sensitivity of the impact scores to the different value functions and of the weighting factors.
4.2.3 Results
4.2.3.1 Case contextualization
The approach is applied to a fictive case study, where four different potential biodiesel suppliers are compared (biodiesel from soy produced in Argentina – product system 1; biodiesel from palm oil produced in Malaysia – product system 2; biodiesel from soy produced in the United States – product system 3; and biodiesel from used oils produced in Québec – product system 4).
In this hypothetical case, an SLCA profile of each product is calculated to assess the social performances of the respective supply chains, as per Figure 4.5. The use stage is not included because it is considered the same for all products. The functional unit is defined as 1 litre of biodiesel available for use.
Figure 124.5 – System boundaries for the social life cycle assessment of a biodiesel supplier 4.2.3.2 Qualitative assessment
We selected eleven subcategories grouped into three stakeholder dimensions (workers, society and local community). The workers stakeholder includes seven subcategory indicators; the local community and society each have two subcategory indicators (Figure 4.6).
Figure 134.6 – Subcategory indicators selected for the case study
For each subcategory indicator, four classification levels were defined, as per the PRP technique (Figure 4.2). Table 4.1 presents an example for the health and security subcategory indicator.
Table 44.1 – Example of classification level definitions for the health and security subcategory indicator (T2)
Health and security subcategory indicator
Description Comparison of the rate of labor accidents in the organization last year and the average rate of the country where the company is located Performance
reference point (PRP)
Average rate of the labor accidents of the country where the company is located in the year 2010 (%)
Classification levels
A The labor accident rate in the company is lower than the average rate in the country. B The labor accident rate in the company is the same as the
average rate in the country. C
The company has a higher rate of labor accidents than the country’s average or the company does not document this indicator and the workers are organized in labor unions. D
The company has a higher rate of labor accidents than the country’s average or the company does not document this indicator and the workers are not organized in labor unions.
For forced labour and child labour, however, we did not follow the PRP and rather considered the SHDB (Social Hotspot Database) indicator. It must be used when no information on the company is available or when it is difficult to obtain specific or reliable data, as suggested by Hosseinijou et al. (2014).
4.2.3.3 Value function
The customized value function converts the qualitative social profile (expressed by the ordinal A- D classification levels) into a cardinal performance. For each subcategory indicator, the scores are defined as an average of a group of three SLCA experts with an extensive experience in SLCA applications.
In Figure 4.7, the customized value function of each indicator presents a different shape. Linear and convex shapes with different scores are obtained for each classification level. The scores are so diverse because the group of SLCA experts placed the classification levels for each indicator at different positions, meaning that the classification levels are valued in different ways depending of the type of the indicator. The position of the classification levels does not influence the relative importance of each subcategory indicator.
Figure 144.7 – Customized value function for each subcategory indicator in Figure 4.6
The quantitative social performance scores of the four potential suppliers are provided in Figure 4.8. The results are presented by life cycle stage.
Supplier 1 Supplier 2 Supplier 3 Supplier 4
Figure 154.8 – Quantitative assessment results obtained from the customized value function for each subcategory indicator in Figure 4.6
Raw Material Production Biodiesel Production Mixing and Stock
4.2.3.4 Aggregation
At the first level, the performances of each subcategory indicator across all life cycle stages of the study are aggregated. Aggregation factors specific to each potential supplier are calculated